I'm a few months late, but I realized that this specific video is just a slight modification of google's "Let’s Write a Decision Tree Classifier from Scratch".
Can't believe I'm just finding this channel! I've been working as a DS for a few years now, but my company is finally delving into ML workflows, and I need a refresher from my undergrad. Turns out, these videos are even more useful than said undergrad! Thanks, nerd. I'll be tying this into other courses as explainers, great videos
For anyone wondering. this is not ID3 because ID3 only works for data sets with categorical attributes, its neither C4.5 because that one generates possibly a general tree and handles both numeric and categorical attributes, and here the algorithm presented generates a binary tree, most close to CART, but CART its for regression and classification, so the guy developed his own algorithm
It's a great tutorial. I have always been struggling to understand these topics. And, now I feel I discovered the right path to get into it. Please keep it up. In addition, it would be much meaningful if the codes you wrote are available for download. Many thanks!
Thanks! In your previous video it wasn't explained how you use information gain to subsequently calculate the best conditions in each node. This explains it very well: recursion!
This is so great...! Just the right video to get me started with Decision Tree Classifier..!!! Keep going, would like to see more videos on other ML algorithms...!!!
Hello from the future, Normalized Nerd! These videos are excellent and incredibly straightforward. Thank you for your contribution and sharing of your knowledge
Imagine paying every month over 100 euros to learn a course based on AI and 30 mins on your vids I've already learned significantly more then I have 2 and a half months with those teachers... Sad world where the teachers hired (A person pretty much directly responsible for future generations) are absolute garbage and 100% don't deserve their pay check with the effort and enthusiasm they put in class. Regarthless, your vids are trully amazing and easy to follow! Good job man and hope these positive comments made your work worth it :) I know it did to me
Hi Great Video! Just wanted to quickly ask if this was based on the zero rule algorithm or CART algorithm. This scratch implementation of decision tree differs from the sklearn's decision tree on my dataset. This implementation trained on my dataset gives me 78% accuracy whereas sklearn's decision tree gives me 95% accuracy. Any ideas what the difference between the 2 implementations could be?
Can you tell me what is the motive of coding decision tree from scratch. Well, we can use the sklearn algo. for it. But, did you do it for just upskilling in python Or, The coded decision tree is better than the sklearn library decision tree. Because, I think the basic algo. of sklearn is also based on the recursive pattern for creating nodes. So, please tell me, what are the benefits of making our own decision tree
I would have loved if you could have gone through the code in more details e.g. explain more thoroughly what each line does and also include an arrow to show in which line of code you are when explaining it. It owuld make it a lot easier for beginners like me to understand your video.
Hey there.. just wanted to ask you if there are more than two attributes, will Machine Learning check information gain from each attribute for every value of attribute..
I am currently studying an AI engineering career and I am certainly worried, because I don´t know how I would solve this if it wasn´t for this video or ChatGPT. What would you recommend me to do to get confidence in programming things like these?
for Node, it seems to simply identify the depth of the current node. in 'get_best_split' it's an index for the loop as it iterates through the range(num_features).
@@secondarypemail7181 It's the index of 'range(num_features)' from the 'get_best_split'. This is passed as a return to the calling function, 'build_tree', just before it's used as an argument in the call to 'Node()' in the __init__ if you look at the 'print_tree' function, it's used there in the print statements to show 'X_' value, as well as in the 'split' function to ensure the value is below the threshold. the range is 4, which gives us the 0-3 we see in the printout of the tree. try plugging it print statements throughout to see where the value changes and to what. I hope that helped.
Good video but a bit hard to follow sometimes on phone when the code is not pointed out with a cursor, cant see exactly what line you are referring to sometimes
Hey! I can easily say that your channel will be worldwide sooner or later! Never compromise from the quality. Great job.
Thanks a lot mate :D :D
I'm a few months late, but I realized that this specific video is just a slight modification of google's "Let’s Write a Decision Tree Classifier from Scratch".
There is just a WOW factor. I had been looking for explanation and implementation from the scratch in this way. Thumbs up to you. Thanks!
Glad to hear that! :D
Can't believe I'm just finding this channel!
I've been working as a DS for a few years now, but my company is finally delving into ML workflows, and I need a refresher from my undergrad. Turns out, these videos are even more useful than said undergrad!
Thanks, nerd. I'll be tying this into other courses as explainers, great videos
This is really awesome!!! Stay blessed and keep producing these great learning videos!
Great work! Thanks for sharing. As you said it "So satisfying to see a model coded from the scratch perform so well" :)
Wow, your videos are extremely high quality
Normalized Nerd please come back we love ur videos !
You put a lot of work in this video, thank you! Subscribed :)
Thanks for the sub!
For anyone wondering. this is not ID3 because ID3 only works for data sets with categorical attributes, its neither C4.5 because that one generates possibly a general tree and handles both numeric and categorical attributes, and here the algorithm presented generates a binary tree, most close to CART, but CART its for regression and classification, so the guy developed his own algorithm
It's a great tutorial. I have always been struggling to understand these topics. And, now I feel I discovered the right path to get into it. Please keep it up. In addition, it would be much meaningful if the codes you wrote are available for download. Many thanks!
This was such a thorough yet approachable explanation. Thanks!
Thanks! In your previous video it wasn't explained how you use information gain to subsequently calculate the best conditions in each node. This explains it very well: recursion!
This is so great...! Just the right video to get me started with Decision Tree Classifier..!!! Keep going, would like to see more videos on other ML algorithms...!!!
For sure!
I enjoyed the code walk through. Nice job. Thank you!
Fantastic video, just like the previous one. Great teacher. Thanks.
Great explanation. I needed the Decision Tree for my research.
you are amazing! hope your channel goes along and more and more popular!
Hello from the future, Normalized Nerd! These videos are excellent and incredibly straightforward. Thank you for your contribution and sharing of your knowledge
Glad you like them!
Wow!! I had been looking for this!
That's great! Keep supporting :D
I ask Chat GPT about something and it broad me here. Man, thanks for this guy and chat GPT, now I got a purpose to do.
This is a great playlist! Keep up the good work
Excellent explanation. The coding is made so simple with your perfect explanation. Thank you :)
Thanks for sharing! I can't wait to start coding!❤
You're the best! Thank you for the great tutorial!
Thank you teacher! I swear you are the best teacher in the world!
Perfect explanation!
TypeError Traceback (most recent call last)
in ()
1 classifier = DecisionTreeClassifier(min_samples_split=3, max_depth=3)
----> 2 classifier.fit(X_train,Y_train)
3 classifier.print_tree()
even i am getting the same error.Have you found the solution for this ?
final part seems a little rush but it's fine.
would appreciate if you explain how the model corelates with the flower dataset
Thanks for the video, this would help me with my btech project.
Thanks bro for clear explanation and well written code
thank you, man!! , You saved it, Nice and easy Explanation.
❤love the content keep it up brother… proud of u
Please continue uploading more information videos like this. 👍👍
Very clear and intuitive! Thanks
Glad it was helpful!
Beautiful explanation. Thank you!
Great videos, keep up the good work
Thanks!! :D
Great work, really appreciate these videos!
Glad you like them!
it is real a well explained video, thanks a lot!!!
Imagine paying every month over 100 euros to learn a course based on AI and 30 mins on your vids I've already learned significantly more then I have 2 and a half months with those teachers... Sad world where the teachers hired (A person pretty much directly responsible for future generations) are absolute garbage and 100% don't deserve their pay check with the effort and enthusiasm they put in class.
Regarthless, your vids are trully amazing and easy to follow! Good job man and hope these positive comments made your work worth it :) I know it did to me
your videos are awesome bro!!!
It is really a great piece of work.
Can you show us on how you can calculate the entropy and the gini within your code?
Hi Great Video!
Just wanted to quickly ask if this was based on the zero rule algorithm or CART algorithm. This scratch implementation of decision tree differs from the sklearn's decision tree on my dataset. This implementation trained on my dataset gives me 78% accuracy whereas sklearn's decision tree gives me 95% accuracy. Any ideas what the difference between the 2 implementations could be?
did you use id3 algorithm for the classification ?
Can you tell me what is the motive of coding decision tree from scratch.
Well, we can use the sklearn algo. for it. But, did you do it for just upskilling in python Or,
The coded decision tree is better than the sklearn library decision tree.
Because, I think the basic algo. of sklearn is also based on the recursive pattern for creating nodes.
So, please tell me, what are the benefits of making our own decision tree
it is useful for learning how the current algorithm works
Very informative.. thanks for sharing
Brilliant video!!
Absolutely amazed
This is good stuff, thanks!
You're welcome!
That is great, sr! Thanks a lot!
In 8:02 minute the "get_best_split()" function takes unnecessary "num_samples" parameter.
u r doing ground breaking work!!! code refer kru kya??
Great Job sir. Very usefull video
So nice of you
Thanks, Nice tutorial
This guy is 3blue1brown but for ML. Thank you!
The best tutorial👍
Thank you..Well explained..
AMAZING, thanks.
Indians at their Best!!! Thanks for such a great content bro!
I would have loved if you could have gone through the code in more details e.g. explain more thoroughly what each line does and also include an arrow to show in which line of code you are when explaining it. It owuld make it a lot easier for beginners like me to understand your video.
Thank you for this video
Hey there.. just wanted to ask you if there are more than two attributes, will Machine Learning check information gain from each attribute for every value of attribute..
Gracias Mil !!! Este video se complementa muy bien con el anterior. De otro lado, tienes algún tutorial sobre Regression Tree?
How do you do those visualizations? They are cool
well explained! Thank you!
Glad to hear that!
hey, is this for C4.5, ID3, or Hunt's algoritm? thx
LOVE THIS
Can I use this code for different values of max_depth ?? Actually I don't want to do any pruning here
I am currently studying an AI engineering career and I am certainly worried, because I don´t know how I would solve this if it wasn´t for this video or ChatGPT. What would you recommend me to do to get confidence in programming things like these?
Can I choose 'entropy' for information gain in this model.
Awesome, man!
Hi, thank you very much for this video. Is there a way to turn the tree we obtain into a more readable graphic representation?
so what does X_2
Thank you for this tutorial
I have a question
Why the program give me
Invalid DecisionTreeClassifier
Is it C4.5 decision tree sir?
in the build tree function there is the line:
X, Y = dataset[: , :-1], dataset[:, - 1]
what do they mean ?
dataset[rows, columns], ":" means all of them so dataset[:,:-1] takes al the rows and all the columns except the last one (index -1).
Very helpful video
@Normalized Nerd how even after using the same code my accuracy is coming 100%?
Greate explaination!!
Thanks!!
Very nice!
I really like it
how to vizualise the above code using graphviz
Good video and the animation is similar to 3blues 1brown, how did you make this video animation I wonder.
I used Manim (an opensource python library) created by 3blue1brown himself!
@@NormalizedNerd yes it is possible but have to write a lot of code
where do i run these commands
Can anyone help me out ,I'm getting error like couldn't able to convert int value into float from datasheet
thank you !
thank you so much
thank you buddy
Can some explain what is feature_index?
for Node, it seems to simply identify the depth of the current node.
in 'get_best_split' it's an index for the loop as it iterates through the range(num_features).
@@alfraelich Okay the the min depth value is given to to the feature index?
@@secondarypemail7181 It's the index of 'range(num_features)' from the 'get_best_split'. This is passed as a return to the calling function, 'build_tree', just before it's used as an argument in the call to 'Node()' in the __init__
if you look at the 'print_tree' function, it's used there in the print statements to show 'X_' value, as well as in the 'split' function to ensure the value is below the threshold. the range is 4, which gives us the 0-3 we see in the printout of the tree.
try plugging it print statements throughout to see where the value changes and to what.
I hope that helped.
Next time please use something besides the iris dataset, if you can. Thanks for the video though.
Well, this was the first time I touched iris in this channel haha...will use more diverse datasets
Thank you but my accuracy_score is 0.36.. :(
awesome
13:48
We need you to trace the code line by line by hand with real data using a very small data set to truly understand what is happening
Pandas is just for the data set. He is doing it from scratch.
Good explanation of the Decision Tree but '__init__' is not a constructor.
Yes, it's not. Mentioned it for the beginners.
Value is pronounced with a /y/ sound, /v ae l y uw/, not as you pronounce it /v ae l uw/.
Bhai tumi Bangali?
Ha :)
bhai code to dedo yar
Hello people from the future 😎
What do you mean
Good video but a bit hard to follow sometimes on phone when the code is not pointed out with a cursor, cant see exactly what line you are referring to sometimes
esrnbarenfarnfbilnrnrr spamming kr rha hu
You save my life, thank you